There is a company near you that looks exactly like your best customer. Same size, same problems, same budget. They have never heard of you, and no vendor sells you that list. The lists you can buy are sorted by industry code and zip code, and your best customer is not an industry code. The list you actually want has to be researched into existence.
In this episode I do exactly that, on screen, start to finish.
It starts with a short prompt, not a persona workshop. I name one real customer, ask Claude to interview me with three questions one at a time, and have it write a one paragraph ideal customer profile in plain English before we use it for anything. The whole definition step takes a couple of minutes, and you watch me answer the questions the same way you would out loud in a hallway.
Then comes the research prompt, and this is where the guardrails matter. Ten companies within 100 miles of San Antonio. Nothing publicly traded. For each one: name, city, website, what they do, rough size, why they fit, and one growth signal with a date on it. No dated signal, no listing. Never guess a name. First pass in a few minutes, then keep digging and rank the final list by strength of fit, because eight strong fits beat ten padded ones.
While it works, it runs the same trail you would run by hand: company sites, directories, news, job postings, the Inc. 5000. It just does not get bored and wander off to Reddit. The first pass came back in a few minutes with named companies, dated signals, and a reason for each one, plus a note on what it could not confirm. A human researcher would have billed a day or two for the same output.
One honest limit: Claude is a researcher, not a contact database. You will not get reliably verified direct emails and phone lines out of this. You will get real companies with real reasons, and the phone number and contact form on their website work fine. Connecting this to a contact database is a future episode.
The point of the whole exercise: a bought list without judgment is spam ammunition. A researched list is a list of potential customers. Run it on your own market, argue with the output, sharpen the profile, and feed the result into the outreach machine you already have.